CN103686345A - Video content comparing method based on digital signal processor - Google Patents
Video content comparing method based on digital signal processor Download PDFInfo
- Publication number
- CN103686345A CN103686345A CN201310701259.4A CN201310701259A CN103686345A CN 103686345 A CN103686345 A CN 103686345A CN 201310701259 A CN201310701259 A CN 201310701259A CN 103686345 A CN103686345 A CN 103686345A
- Authority
- CN
- China
- Prior art keywords
- data
- video data
- characteristic vector
- normal video
- video
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
The invention discloses a video content comparing method based on a digital signal processor. The method comprises the steps that single-channel video bitstream data and standard video bitstream data are forwarded to DSP modules according to the load balanced scheduling strategy, and the DSP modules obtain the key frame image sequence of the video data and the key frame image sequence of the standard video data; the DSP modules extract the characteristics of the key frame image sequence of the video data and the key frame image sequence of the standard video data according to the improved SIFT algorithm and the TOM algorithm to obtain the feature vector sequence of the video data and the feature vector sequence of the standard video data to determine the synchronous video data and standard video data, and if the similarity of the feature vector sequence of the video data and the feature vector sequence of the standard video data is larger than the preset similarity threshold value, the contents are identical, wherein the video data and the standard video data are synchronous. The method has the advantages that the data are processed in real time, the volume is large, the size is small, the power consumption is low and labor cost is low, and is used for solving the problems that the video processing cannot meet the requirements for real-time performance, the large volume, the small size and the low power consumption.
Description
Technical field
The present invention relates to video detection technology field, particularly relate to a kind of video content comparison method based on digital signal processor.
Background technology
Along with the development of network and the communication technology, Video Applications has been generalized to every field, and it is made and propagates also without technical threshold, causes video capacity surge, content to spread unchecked.Meanwhile, as culture and ideology communication media, video malicious attack event takes place frequently, and the requirement of video regulation technique is further improved.In many supervision and detection method, video content supervision can be carried out depth detection to video, is regulatory measure the most thoroughly in video supervision field.
Yet the realization of video content supervision faces many problems, main cause has two:
One, video have base requirement when strict, and this just requires video processor to have higher processing capability in real time, can synchronously process in strict accordance with the time base of video.But the video content analysis based on x86 IA frame serverPC is difficult to requirement of real time.
Two, video has the feature that data volume is large, and in the massive video data epoch, the video content analysis equipment of prior art is difficult to meet the content analysis to multitude of video data.How under equal supervision capacity situation, reducing huge Content analysis apparatus volume, thereby develop the video content analysis equipment of large capacity, small size, low-power consumption, is also the problem that content-based methods of video analyses will solve.
In view of the existence of above two subject matters, the research of video content analysis still rests on theoretical level and experimental stage.
Summary of the invention
The technical problem to be solved in the present invention is to utilize MIMD multiple-instruction-stream multiple-data stream (MIMD) to process structure and the extensive video data of loose coupling storage organization parallel processing, reduces the time complexity of video algorithm.The mode that simultaneously adopts multichannel code stream load balance scheduling strategy to combine with parallel video Processing Algorithm guarantees the requirement of real-time of video content analysis, and multiple-path multiple-core DSP is applied in real-time video content analysis comparison.In order to solve the video content analysis equipment of prior art, can not meet the problem that real-time, large capacity, small size, low-power consumption require.
For solving the problems of the technologies described above, the invention provides a kind of video content comparison method based on digital signal processor, the method comprises the following steps: the multichannel video bit stream data and the multichannel normal video bit stream data that receive Internet Transmission; The channel information demultiplexing carrying according to multichannel video bit stream data and multichannel normal video bit stream data is single channel video codeword data stream and single channel normal video bit stream data, and according to load balance scheduling strategy to each digital signal processor module forwards single channel video codeword data stream and single channel normal video bit stream data; Video codeword data stream and normal video bit stream data that each digital signal processor module decoding receives, and obtain the key frame images sequence of the video data in video codeword data stream and the key frame images sequence of the normal video data in normal video bit stream data; Described digital signal processor module is extracted the feature of the key frame images sequence of video data and normal video data, to obtain the characteristic vector sequence of video data and the characteristic vector sequence of normal video data; The feature of the video data extracting based on described digital signal processor module and the key frame images sequence of normal video data, determines synchronous video data and normal video data; In synchronous video data and normal video data, calculate the similarity of the characteristic vector sequence of video data and the characteristic vector sequence of normal video data; If described similarity is greater than default similarity threshold, video codeword data stream is identical with the content of normal video bit stream data; If described similarity is less than or equal to default similarity threshold, the content of video codeword data stream and normal video bit stream data is not identical.
Wherein, load balance scheduling strategy is to each digital signal processor module forwards single channel video codeword data stream and single channel normal video bit stream data, comprise: set in advance the priority of each digital signal processor module, for state parameter is single channel video codeword data stream and single channel normal video bit stream data described in digital signal processor module forwards idle and that priority is the highest; When wherein the data volume of digital signal processor resume module is less than predetermined value, state parameter is for idle.
Wherein, described digital signal processor module is extracted the feature of the key frame images sequence of video data and normal video data, comprising: in described key frame images sequence, read a frame image data, as current image date; From spatial domain, adopt improved single size SIFT algorithm to extract characteristic point in described current image date the characteristic vector of calculated characteristics point; Wherein, described improved single size SIFT algorithm refers to the multiple dimensioned processing of only carrying out single dimension D oG view data when extract minutiae; Described current image date is divided into upper left, upper right, lower-left, region, 4 of bottom rights; In time domain, from described image sequence, read front 3 frame image datas of current image date, and obtain the characteristic vector of the characteristic point of described front 3 frame image datas; The characteristic vector of calculating described current image date and described front 3 frame image datas each corresponding region in described 4 regions is respectively in the characteristic vector average of 8 directions, to obtain the characteristic vector of described current image date.
Wherein, described digital signal processor module is utilized improved single size SIFT algorithm and TOM algorithm, extract the feature of the key frame images sequence of video data and normal video data, to obtain the characteristic vector sequence of video data and the characteristic vector sequence of normal video data, comprising: the frame in the image sequence of video data or the characteristic vector of multiple image data form the characteristic vector sequence of video data; A frame in the image sequence of normal video data or the characteristic vector of multiple image data form the characteristic vector sequence of normal video data.
Wherein, based on described digital signal processor module, extract the feature of the key frame images sequence of video data and normal video data, determine synchronous video data and normal video data, comprise: in key frame images sequence, read a frame image data, and calculate the characteristic point of this view data and the characteristic vector of characteristic point; In described view data, delimit the retaining zone of preset range; Described retaining zone is divided into 4 parts, in each part, calculate respectively the average of 8 characteristic vectors on predetermined direction, as the retaining zone characteristic vector of described view data, to utilize retaining zone characteristic vector to determine synchronous video data and normal video data.
Wherein, based on described digital signal processor module, extract the feature of the key frame images sequence of video data and normal video data, determine synchronous video data and normal video data, comprising: the retaining zone characteristic vector of the view data being comprised by video data forms video data retaining zone characteristic vector sequence; The frame that selection standard video data comprises or the retaining zone characteristic vector of multiple image data form normal video data retaining zone characteristic vector sequence, using as synchronous window; Described synchronous window is slided in described video data retaining zone characteristic vector sequence, to search the part identical with described synchronous window in video data retaining zone characteristic vector sequence; If find the part identical with described synchronous window described video data and described normal video data synchronous.
Wherein, described synchronous window is slided in described video data retaining zone characteristic vector sequence, to search the part identical with synchronous window in video data retaining zone characteristic vector sequence, comprise: when searching the time of the part identical with described synchronous window, surpass after the scheduled time, output alarm information, to report to the police.
Wherein, in synchronous video data and normal video data, calculate the similarity of the characteristic vector sequence of video data and the characteristic vector sequence of normal video data, also comprise: if described similarity is less than or equal to default similarity threshold, output alarm information, to report to the police.
Beneficial effect of the present invention is as follows:
The present invention utilizes MIMD multiple-instruction-stream multiple-data stream (MIMD) to process structure and the extensive video data of loose coupling storage organization parallel processing, reduces the time complexity of video algorithm.The mode that simultaneously adopts multichannel code stream load balance scheduling strategy to combine with parallel video Processing Algorithm guarantees the requirement of real-time of video content analysis, and multiple-path multiple-core DSP is applied in real-time video content analysis comparison, improve the disposal ability of video content analysis equipment, make to process and there is real-time, so make that video content analysis place capacity is large, volume is little, low in energy consumption.The present invention can reduce cost of labor, improves man efficiency.
Accompanying drawing explanation
Fig. 1 is DSP media processing board hardware configuration schematic diagram according to an embodiment of the invention;
Fig. 2 is the schematic diagram of the video content comparison method based on digital signal processor according to an embodiment of the invention;
Fig. 3 is the flow chart of the video content comparison method based on digital signal processor according to an embodiment of the invention;
Fig. 4 receives according to an embodiment of the invention and dispatches step schematic diagram;
Fig. 5 is characteristic extraction procedure schematic diagram according to an embodiment of the invention; And
Fig. 6 is synchronizing step schematic diagram according to an embodiment of the invention.
Embodiment
In order to solve in the video content analysis of prior art, due to the application of video processnig algorithms, restrictions such as the disposal ability of processor and hardware integration metric moulds, video content analysis equipment can not meet the problem of real-time, large capacity, small size, low-power consumption requirement, the invention provides a kind of video content comparison method based on digital signal processor.Below in conjunction with accompanying drawing, the present invention is further elaborated.Should be appreciated that specific embodiment described herein, only in order to explain the present invention, does not limit the present invention.
The present invention is based on multipath high-speed parallel digital signal processor (DigitalSignalProcessing, DSP), employing is applicable to the analyses and comparison technology of the content-based feature extraction of parallel organization, and guarantee from many aspects such as system, algorithm, optimizations accuracy and the real-time that video content is compared, to realize the automatic real-time monitoring to multi-channel video content.
Wherein, digital signal processor can be the media processing board of hardware, and as shown in Figure 1, Fig. 1 is DSP media processing board hardware configuration schematic diagram according to an embodiment of the invention.
This board can comprise optical switch module, control and management module, ethernet switching module, interrupt module, data exchange module, memory module and one or more DSP module (Fig. 1 middle finger schematically illustrates two DSP modules).Wherein, optical switch module and ethernet switching module can be for by light Fabric Interface and Ethernet interface receiver, video bit stream data and normal video bit stream datas and be transmitted to corresponding DSP module, to guarantee height degree of the gulping down rate of data.Control and management module is that the center of board is controlled and administrative unit, can enter corresponding DSP module for video codeword data stream and the normal video bit stream data of scheduling optical switch module and ethernet switching module, and schedule information is stored in memory module.The task that interrupt module can carried out for suspending DSP module.Data exchange module can be injected into DSP module for the schedule information that control and management module is generated, so that DSP module is carried out corresponding task, and DSP module is transferred to control and management module by data exchange module by warning message.
One or more DSP modules are core function unit, can be for video codeword data stream being carried out to the analysis of the video content based on normal video bit stream data and comparing.Further, also comprise DSP Status Monitor (not shown in figure 1) in each DSP module, this DSP Status Monitor can monitor the state parameter of DSP module, generates state parameter and feeds back to control and management module by data exchange module.
Wherein video codeword data stream can be video, as a certain video APP(Application, application) TV programme play.Normal video bit stream data can be the video from the default network address, as the TV programme of Chinese Network TV Station broadcasting.In video codeword data stream and normal video bit stream data, carry channel information, the TV programme that the TV programme that APP can be play like this and Chinese Network TV Station play is carried out video content contrast, with the video that APP is play, monitor, prevent the situation generation that malice video is attacked.
Fig. 2 is the schematic diagram of the video content comparison method based on digital signal processor according to an embodiment of the invention.Receiving and dispatching in step, control and management module receives real-time multichannel video codeword data stream from light Fabric Interface or Ethernet interface, to this multi-channel video bit stream data demultiplexing, and according to the channel information of video codeword data stream, as: CCTV1, CCTV2, is separated into one or more single channel video codeword data streams by video codeword data stream.In like manner, can go out one or more single channel normal video bit stream datas by demultiplexing.Control and management module can utilize load balance scheduling strategy to dispatch, to send single channel video data and single channel normal video bit stream data to corresponding DSP module.Stateful Inspection and feedback step refer to that the DSP Status Monitor in DSP module feeds back the current state parameter of DSP module to control and management module, as: the states such as DSP module is busy, not busy, this parameter can provide reference for load balance scheduling strategy.In each DSP module, whether the single channel video codeword data stream receiving for this DSP module and single channel normal video bit stream data pass through respectively preliminary treatment (step S320), feature extraction (step S330), synchronous (step S340) and aspect ratio to (step S350), and then analyze video codeword data stream and conform to the content of normal video bit stream data.If exist the discrepancy can output alarm information, report to the police implementing.
Below in conjunction with Fig. 3-Fig. 6, the content in Fig. 2 and the step relating to are elaborated.
Fig. 3 is the flow chart of the video content comparison method based on digital signal processor according to an embodiment of the invention.
Step S310 is for receiving and scheduling step, in this step, board receives multichannel video bit stream data and the multichannel normal video bit stream data of Internet Transmission, the channel information carrying according to multichannel video bit stream data and multichannel normal video bit stream data, demultiplexing is one or more single channel video codeword data streams and single channel normal video bit stream data, and according to load balance scheduling strategy to each DSP module forwards single channel video codeword data stream and single channel normal video bit stream data.
Load balance scheduling strategy refers to according to the state parameter of the channel information in the video codeword data stream receiving and each DSP Status Monitor feedback, makes the treating capacity of each DSP module impartial.Further, when the data volume of DSP resume module is less than predetermined value, its state parameter is idle; Otherwise its state parameter is for busy; Set in advance the priority of each DSP module; Particularly, can be each DSP preset numbers, each numbering represents a priority level.During forwarding data, for state parameter is DSP module forwards single channel video codeword data stream and single channel normal video bit stream data idle and that priority is the highest, that is: in state parameter is idle each DSP module, by numbering the highest DSP resume module data of priority, thereby realize the load balancing of each DSP module.
Reception and scheduling step schematic diagram with reference to the one embodiment of the invention shown in figure 4.Control and management module is responsible for receiving and scheduling step, and this control and management module can comprise that network receives buffer memory, channel separator, task dispatcher, DSP state collector, dsp controller.Particularly, the DSP Status Monitor in DSP module is responsible for to the state parameter of control and management module feedback DPS module, to assist control and management module to complete the scheduling to video codeword data stream and normal video bit stream data.Particularly, can put into network reception buffer memory by receive multichannel video bit stream data and normal video bit stream data from network.Multichannel video bit stream data and the normal video bit stream data by channel separator, being responsible for network to receive in buffer memory are separated into respectively single channel video codeword data stream and normal video bit stream data.The state parameter that DSP state collector is carried out to the one or more DSP module design task of DSP Status Monitor request.DSP Status Monitor is to the state parameter of DSP state collector feedback DSP module.Wherein, the state parameter of DSP module can comprise the state parameters such as DSP module is idle, busy.DSP state collector is delivered to task dispatcher by the state parameter of DSP module.Task dispatcher produces schedule information according to load balance scheduling strategy, and this schedule information is sent to dsp controller, controls the operation task of DSP module, as received normal video bit stream data and processing.
In this step, control and management module can be removed the redundant information of packing because adapting to transmission in video codeword data stream and normal video bit stream data, as: the information such as the object network address, source network address, and then obtain video data in video codeword data stream and the normal video data in normal video bit stream data, only video data and normal video data are transmitted to corresponding DSP module according to load balance scheduling strategy.
Each DSP module can parallel processing three road video codeword data streams and/or normal video bit stream data.The present invention, for video codeword data stream and normal video bit stream data are carried out to content comparison, can make each DSP module parallel processing two-path video bit stream data and the accurate video codeword data stream of a road sign, as shown in Figure 2.
Following steps S320-step S350 all carries out in DSP module.
Step S320 is pre-treatment step, video codeword data stream and normal video bit stream data that each digital signal processor module decoding receives, and obtain the key frame images sequence of video data in video codeword data stream and the key frame images sequence of the normal video data in normal video bit stream data.This step can reduce data volume to be processed.
In the two field picture that video data and normal video packet contain, can comprise the frame information of key frame and non-key frame, according to the frame information in two field picture, can filter out the key frame images of video data and normal video data, and delete non-key two field picture.Further, can the key frame images in video data form the image sequence of video data, the key frame images in normal video data forms the image sequence of normal video data.
Generally speaking, in video data and normal video data, in every 24 two field pictures, only comprise 2 frame key frame images.Therefore,, in a large amount of videos, the gray level image of use key frame can effectively reduce the data volume of data processing, improves treatment effeciency.
In one embodiment, key frame images is converted to gray level image, only utilizes the gray level image of key frame to carry out follow-up processing, reduction closer data volume, improved treatment effeciency.
Step S330 is characteristic extraction step, utilize improved single size SIFT(Scale-invariant feature transform, the conversion of yardstick invariant features) algorithm and TOM(Topic & Occurrence-oriented Merging) algorithm, extract the feature of video data and normal video data, to obtain the characteristic vector sequence of video data and the characteristic vector sequence of normal video data.Wherein, this feature can comprise the characteristic vector of characteristic point and characteristic point.
In order to adapt to the real-time comparison requirement based on video content.The mode that the present invention adopts improved single size SIFT algorithm to combine with TOM algorithm is extracted feature and is made up the impact of characteristic vector dimension-reduction treatment on characteristic area calibration.
Wherein, improved single size SIFT algorithm refers to and when extract minutiae, only carries out single dimension D oG(Difference of Gaussian) the multiple dimensioned processing of view data, do not carry out the processing of many sizes (or claiming multistage) view data, without sub-sampling process; To guarantee in the metastable situation of characteristic point, improve real-time.Secondly, because feature extraction algorithm need to be to rotating sensitivity, so there is no the concept of principal direction, auxiliary direction, all images are calculated characteristics vector on assigned direction only.
TOM algorithm is to utilize the spatial feature of video and the feature extraction algorithm of temporal signatures simultaneously.Spatial feature refers to take the feature that space is reference quantity.Temporal signatures refers to take the feature that the time is reference quantity.
This characteristic extraction step schematic diagram of characteristic extraction procedure according to an embodiment of the invention as shown in Figure 5.Utilize TOM algorithm, in the image sequence of key frame, read a frame image data, from spatial domain, adopt improved single size SIFT algorithm to extract the characteristic point in view data, the characteristic vector of calculated characteristics point.View data is divided into upper left, upper right, lower-left, region, 4 of bottom rights.Calculate characteristic point in each region characteristic vector average on 8 predetermined different directions, thereby can obtain 32 dimensional feature vectors (4 regions of 8 direction *) in the spatial domain of this frame image data.In time domain, from image sequence, read front 3 frame image datas of current image date, and obtain the characteristic vector of this 3 frame image data.The characteristic vector of calculating 4 view data corresponding regions is respectively for example, in the characteristic vector average of 8 directions: the corresponding top left region of top left region.So far obtain the frame image features vector that this two field picture time domain and spatial domain combine.
A frame in the image sequence of video data or the characteristic vector of multiple image data form the characteristic vector sequence of video data.A frame in the image sequence of normal video data or the characteristic vector of multiple image data form the characteristic vector sequence of normal video data.
Adopt which to extract feature, recodification to video codeword data stream is insensitive, and comparatively responsive to the variation of the sequential of frame of video (sequencing of frame of video), duration (duration of certain section of video), content (brightness, colourity, the part that comprise image), so can effectively compare out the variation of video content.
And, utilize improved single size SIFT algorithm and the TOM algorithm mode of combining to extract the feature used time and can reach ms level, improved real-time.
Step S340 is synchronizing step, extracts the feature of video data and normal video data based on described DSP module, determines synchronous video data and normal video data.
After the feature of extracting every frame image data, can enter synchronizing step, make to obtain parallel the carrying out of the characteristic vector sequence of video data and the characteristic vector sequence of normal video data and synchronizing step, to improve the treatment effeciency of content comparison.
In order to consider real time problems, and reduce operand, using the middle part of every two field picture as between synchronous comparison area.Particularly, in view data, find characteristic point and calculate the characteristic vector of each characteristic point, in view data, delimit the retaining zone of preset range, in this retaining zone, adopt the mode similar to characteristic extraction step S330, utilize improved single size SIFT algorithm and TOM algorithm, calculate the characteristic vector in retaining zone.
Fig. 6 is synchronizing step schematic diagram according to an embodiment of the invention.In the image sequence of key frame, read a frame image data, find the characteristic point of this view data and the characteristic vector of characteristic point, in this view data, delimit the retaining zone of preset range, retaining zone is divided into 4 parts, in each part, calculate respectively the average of 8 characteristic vectors on predetermined direction, as the retaining zone characteristic vector of view data, to utilize retaining zone characteristic vector to determine synchronous video data and normal video data.
The retaining zone characteristic vector of the view data being comprised by video data forms video data retaining zone characteristic vector sequence.
The frame that selection standard video data comprises or the retaining zone characteristic vector of multiple image data form normal video data retaining zone characteristic vector sequence, using as synchronous window.For example, choose the characteristic vector formation synchronous window of 4 frame image datas continuous in time domain.
Synchronous window is slided in video data retaining zone characteristic vector sequence, in video data retaining zone characteristic vector sequence, search the part identical with synchronous window.If find identical part, think that video data synchronizes with normal video data, can perform step S350.If there is not identical part, do not utilize synchronous window to continue to search identical part.When searching the time of the part identical with synchronous window, surpass after the scheduled time, can output alarm information, to report to the police, to be checked finding after same section, input stop alarm information.
Step S350 be aspect ratio to step, in synchronous video data and normal video data, calculate the similarity of video data and normal video data.
When video data and normal video data are when synchronous, the image sequence of the image sequence of the key frame images composition video data in video data and the composition of the key frame images in normal video data normal video data can be carried out to similarity comparison, to obtain similarity.When this similarity is greater than default similarity threshold, video data and normal video data is identical, and video codeword data stream is identical with the content of normal video bit stream data.When this similarity is less than or equal to default similarity threshold, video data and normal video data are not identical, and the content of video codeword data stream and normal video bit stream data is not identical.When video codeword data stream is not identical with normal video bit stream data, output alarm information, to report to the police.
Further, the calculating of this similarity adopts position sensing hash algorithm (LSH, Locality Sensitive Hash).Should be appreciated that similarity algorithm of the present invention is not limited to this.
The present invention is based on MIMD multiple-instruction-stream multiple-data stream (MIMD, MultipleInstructionStreamMultipleDataStream) and process the multidiameter delay DSP module of structure and loose coupling storage organization.Wherein, MIMD processes structure and refers to the structure by array DSP module composition in conjunction with the function of DSP module.Loose coupling storage organization refers to that each the DSP module in a plurality of DSP modules has the memory space of independent use.This design feature can be carried out instruction stream separately in video data stream separately, and then can realize the processing to multi-path video stream.
The present invention adopts the automatic analysis alignment algorithm of the video content features that is suitable for this parallel organization, coordinate control and management module and scheduling strategy thereof, realize individual module all the period of time of the cleer and peaceful 20 tunnel high definition real-time video code streams of 60 road signs is supervised to (single place capacity can reach the cleer and peaceful 60 tunnel high definitions of 180 road signs) automatically, reduce video supervision field because of the malicious attack supervision accident that regulatory measure backwardness causes that takes place frequently, reduce artificial the cost of supervise and control, thereby reduce the cost of supervise and control in the situation that improving supervisory efficiency.The present invention forms material object by theoretical the combining with hardware device of Video processing, can supervise the real-time video code stream of the TV broadcasting system of crucial website, gateway, network node and IP based network, for the field that has all the period of time, large capacity and real-time video supervision to require provides solution.
The present invention is except reduce time complexity and space complexity on algorithm, and make it to be applicable to outside parallel processing structure, also adopt Real-Time Scheduling Police, streamline to combine with divide and conquer, make full use of hardware parallel processing capability, thereby reach the object to large-capacity video data processing and real-time analysis.
The present invention is based on multipath high-speed Parallel DSP, employing is applicable to the content-based feature extraction analyses and comparison technology of parallel organization, and guarantee from a plurality of steps such as system, algorithm, optimization accuracy and the real-time compared, video content analysis method is risen to application height from theoretical aspect, realize the automatic real-time monitoring to multi-channel video content.
Although be example object, the preferred embodiments of the present invention are disclosed, it is also possible those skilled in the art will recognize various improvement, increase and replacement, therefore, scope of the present invention should be not limited to above-described embodiment.
Claims (8)
1. the video content comparison method based on digital signal processor, is characterized in that, said method comprising the steps of:
Receive multichannel video bit stream data and the multichannel normal video bit stream data of Internet Transmission;
The channel information carrying according to multichannel video bit stream data and multichannel normal video bit stream data, demultiplexing is single channel video codeword data stream and single channel normal video bit stream data, and according to load balance scheduling strategy to each digital signal processor module forwards single channel video codeword data stream and single channel normal video bit stream data;
Video codeword data stream and normal video bit stream data that each digital signal processor module decoding receives, obtain the key frame images sequence of the video data in video codeword data stream, and the key frame images sequence of the normal video data in normal video bit stream data;
Described digital signal processor module is extracted the feature of the key frame images sequence of video data and normal video data, to obtain the characteristic vector sequence of video data and the characteristic vector sequence of normal video data;
The feature of the video data extracting based on described digital signal processor module and the key frame images sequence of normal video data, determines synchronous video data and normal video data;
In synchronous video data and normal video data, calculate the similarity of the characteristic vector sequence of video data and the characteristic vector sequence of normal video data;
If described similarity is greater than default similarity threshold, video codeword data stream is identical with the content of normal video bit stream data;
If described similarity is less than or equal to default similarity threshold, the content of video codeword data stream and normal video bit stream data is not identical.
2. method according to claim 1, is characterized in that, according to load balance scheduling strategy, to each digital signal processor module forwards single channel video codeword data stream and single channel normal video bit stream data, comprising:
Set in advance the priority of each digital signal processor module, for state parameter is single channel video codeword data stream and single channel normal video bit stream data described in digital signal processor module forwards idle and that priority is the highest; When wherein the data volume of digital signal processor resume module is less than predetermined value, state parameter is for idle.
3. method according to claim 1, is characterized in that, described digital signal processor module is extracted the feature of the key frame images sequence of video data and normal video data, comprising:
In described key frame images sequence, read a frame image data, as current image date;
From spatial domain, adopt improved single size SIFT algorithm to extract characteristic point in described current image date the characteristic vector of calculated characteristics point; Wherein, described improved single size SIFT algorithm refers to many sizes processing of only carrying out single dimension D oG view data when extract minutiae;
Described current image date is divided into upper left, upper right, lower-left, region, 4 of bottom rights;
In time domain, from described image sequence, read front 3 frame image datas of current image date, and obtain the characteristic vector of the characteristic point of described front 3 frame image datas;
The characteristic vector of calculating described current image date and described front 3 frame image datas each corresponding region in described 4 regions is respectively in the characteristic vector average of 8 directions, to obtain the characteristic vector of described current image date.
4. method according to claim 3, it is characterized in that, described digital signal processor module is extracted the feature of the key frame images sequence of video data and normal video data, to obtain the characteristic vector sequence of video data and the characteristic vector sequence of normal video data, comprising:
A frame in the image sequence of video data or the characteristic vector of multiple image data form the characteristic vector sequence of video data;
A frame in the image sequence of normal video data or the characteristic vector of multiple image data form the characteristic vector sequence of normal video data.
5. method according to claim 1, is characterized in that, extracts the feature of the key frame images sequence of video data and normal video data based on described digital signal processor module, determines synchronous video data and normal video data, comprising:
In key frame images sequence, read a frame image data, and calculate the characteristic point of this view data and the characteristic vector of characteristic point;
In described view data, delimit the retaining zone of preset range;
Described retaining zone is divided into 4 parts, in each part, calculate respectively the average of 8 characteristic vectors on predetermined direction, as the retaining zone characteristic vector of described view data, to utilize retaining zone characteristic vector to determine synchronous video data and normal video data.
6. method according to claim 5, is characterized in that, extracts the feature of the key frame images sequence of video data and normal video data based on described digital signal processor module, determines synchronous video data and normal video data, comprising:
The retaining zone characteristic vector of the view data being comprised by video data forms video data retaining zone characteristic vector sequence;
The frame that selection standard video data comprises or the retaining zone characteristic vector of multiple image data form normal video data retaining zone characteristic vector sequence, using as synchronous window;
Described synchronous window is slided in described video data retaining zone characteristic vector sequence, to search the part identical with described synchronous window in video data retaining zone characteristic vector sequence;
If find the part identical with described synchronous window described video data and described normal video data synchronous.
7. method according to claim 6, is characterized in that, described synchronous window is slided in described video data retaining zone characteristic vector sequence, to search the part identical with synchronous window in video data retaining zone characteristic vector sequence, comprising:
When searching the time of the part identical with described synchronous window, surpass after the scheduled time, output alarm information, to report to the police.
8. method according to claim 1, it is characterized in that, in synchronous video data and normal video data, calculate the similarity of the characteristic vector sequence of video data and the characteristic vector sequence of normal video data, also comprise: if described similarity is less than or equal to default similarity threshold, output alarm information, to report to the police.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310701259.4A CN103686345B (en) | 2013-12-18 | 2013-12-18 | Video content comparing method based on digital signal processor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310701259.4A CN103686345B (en) | 2013-12-18 | 2013-12-18 | Video content comparing method based on digital signal processor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103686345A true CN103686345A (en) | 2014-03-26 |
| CN103686345B CN103686345B (en) | 2017-01-11 |
Family
ID=50322413
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310701259.4A Active CN103686345B (en) | 2013-12-18 | 2013-12-18 | Video content comparing method based on digital signal processor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103686345B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104980764A (en) * | 2014-04-14 | 2015-10-14 | 深圳中兴力维技术有限公司 | Parallel coding/decoding method, device and system based on complexity balance |
| CN104994426A (en) * | 2014-07-07 | 2015-10-21 | Tcl集团股份有限公司 | Method and system of program video recognition |
| CN105072455A (en) * | 2015-08-11 | 2015-11-18 | 精硕世纪科技（北京）有限公司 | Video matching method and device |
| CN105357570A (en) * | 2015-11-03 | 2016-02-24 | 上海熙菱信息技术有限公司 | Video stream analysis method and system based on frame analysis |
| CN107071559A (en) * | 2017-05-11 | 2017-08-18 | 大连动感智慧科技有限公司 | Many video comparison systems based on crucial frame synchronization |
| CN107748569A (en) * | 2017-09-04 | 2018-03-02 | 中国兵器工业计算机应用技术研究所 | Motion control method, device and UAS for unmanned plane |
| CN107944754A (en) * | 2017-12-13 | 2018-04-20 | 泰康保险集团股份有限公司 | Method, apparatus, storage medium and the electronic equipment of rehabilitation performance rating |
| CN108337167A (en) * | 2018-02-11 | 2018-07-27 | 福州大学 | Video multi-channel parallel transmission and distribution method and system based on ant colony algorithm |
| CN110493609A (en) * | 2019-08-07 | 2019-11-22 | 咪咕文化科技有限公司 | Live broadcasting method, terminal and computer readable storage medium |
| CN110738135A (en) * | 2019-09-25 | 2020-01-31 | 艾普工华科技（武汉）有限公司 | worker work step specification visual identification judgment and guidance method and system |
| CN111642004A (en) * | 2020-05-29 | 2020-09-08 | 河南科技大学 | Cooperative communication transmission protocol based on critical data in WBAN |
| CN113516006A (en) * | 2021-04-01 | 2021-10-19 | 广州云硕科技发展有限公司 | Efficient information processing method and system for intelligent robot |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060188020A1 (en) * | 2005-02-24 | 2006-08-24 | Wang Zhicheng L | Statistical content block matching scheme for pre-processing in encoding and transcoding |
| CN101262426A (en) * | 2008-04-28 | 2008-09-10 | 艾诺通信系统（苏州）有限责任公司 | DSP resource management method in IP multimedia system |
| CN101540907A (en) * | 2009-04-28 | 2009-09-23 | 无锡天脉聚源传媒科技有限公司 | Method for auditing safety of internet video |
| CN103430175A (en) * | 2011-01-07 | 2013-12-04 | 阿尔卡特朗讯公司 | Method and apparatus for comparing videos |
-
2013
- 2013-12-18 CN CN201310701259.4A patent/CN103686345B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060188020A1 (en) * | 2005-02-24 | 2006-08-24 | Wang Zhicheng L | Statistical content block matching scheme for pre-processing in encoding and transcoding |
| CN101262426A (en) * | 2008-04-28 | 2008-09-10 | 艾诺通信系统（苏州）有限责任公司 | DSP resource management method in IP multimedia system |
| CN101540907A (en) * | 2009-04-28 | 2009-09-23 | 无锡天脉聚源传媒科技有限公司 | Method for auditing safety of internet video |
| CN103430175A (en) * | 2011-01-07 | 2013-12-04 | 阿尔卡特朗讯公司 | Method and apparatus for comparing videos |
Non-Patent Citations (3)
| Title |
|---|
| NAEL HIRZALLAH ETAL: "A Fast Method to Spot a Video Sequence within a Live Stream", 《JOURNAL OF MULTIMEDIA》 * |
| 杨婧: "基于内容比对的视频检索技术在广播电视安全播出中的应用研究", 《2012中国数字广播电视与网络发展年会》 * |
| 都思丹: "嵌入式多核处理器系统及视频信号处理技术研究进展", 《南京大学学报》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104980764A (en) * | 2014-04-14 | 2015-10-14 | 深圳中兴力维技术有限公司 | Parallel coding/decoding method, device and system based on complexity balance |
| CN104980764B (en) * | 2014-04-14 | 2019-06-21 | 深圳力维智联技术有限公司 | Parallel decoding method, apparatus and system based on complex degree equalization |
| CN104994426A (en) * | 2014-07-07 | 2015-10-21 | Tcl集团股份有限公司 | Method and system of program video recognition |
| CN104994426B (en) * | 2014-07-07 | 2020-07-21 | Tcl科技集团股份有限公司 | Program video identification method and system |
| CN105072455A (en) * | 2015-08-11 | 2015-11-18 | 精硕世纪科技（北京）有限公司 | Video matching method and device |
| CN105357570B (en) * | 2015-11-03 | 2018-10-23 | 上海熙菱信息技术有限公司 | A kind of video flowing analysis method and system based on frame analysis |
| CN105357570A (en) * | 2015-11-03 | 2016-02-24 | 上海熙菱信息技术有限公司 | Video stream analysis method and system based on frame analysis |
| CN107071559A (en) * | 2017-05-11 | 2017-08-18 | 大连动感智慧科技有限公司 | Many video comparison systems based on crucial frame synchronization |
| CN107748569B (en) * | 2017-09-04 | 2021-02-19 | 中国兵器工业计算机应用技术研究所 | Motion control method and device for unmanned aerial vehicle and unmanned aerial vehicle system |
| CN107748569A (en) * | 2017-09-04 | 2018-03-02 | 中国兵器工业计算机应用技术研究所 | Motion control method, device and UAS for unmanned plane |
| CN107944754A (en) * | 2017-12-13 | 2018-04-20 | 泰康保险集团股份有限公司 | Method, apparatus, storage medium and the electronic equipment of rehabilitation performance rating |
| CN108337167A (en) * | 2018-02-11 | 2018-07-27 | 福州大学 | Video multi-channel parallel transmission and distribution method and system based on ant colony algorithm |
| CN108337167B (en) * | 2018-02-11 | 2020-07-31 | 福州大学 | Video multi-channel parallel transmission and distribution method and system based on ant colony algorithm |
| CN110493609A (en) * | 2019-08-07 | 2019-11-22 | 咪咕文化科技有限公司 | Live broadcasting method, terminal and computer readable storage medium |
| CN110493609B (en) * | 2019-08-07 | 2022-02-01 | 咪咕文化科技有限公司 | Live broadcast method, terminal and computer readable storage medium |
| CN110738135A (en) * | 2019-09-25 | 2020-01-31 | 艾普工华科技（武汉）有限公司 | worker work step specification visual identification judgment and guidance method and system |
| CN110738135B (en) * | 2019-09-25 | 2023-06-09 | 艾普工华科技（武汉）有限公司 | Method and system for judging and guiding worker operation step standard visual recognition |
| CN111642004A (en) * | 2020-05-29 | 2020-09-08 | 河南科技大学 | Cooperative communication transmission protocol based on critical data in WBAN |
| CN111642004B (en) * | 2020-05-29 | 2024-02-06 | 河南科技大学 | Cooperative communication transmission protocol based on key data in WBAN |
| CN113516006A (en) * | 2021-04-01 | 2021-10-19 | 广州云硕科技发展有限公司 | Efficient information processing method and system for intelligent robot |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103686345B (en) | 2017-01-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103686345A (en) | Video content comparing method based on digital signal processor | |
| CN107067365A (en) | The embedded real-time video stream processing system of distribution and method based on deep learning | |
| US20210243388A1 (en) | Optimized video review using motion recap images | |
| CN102497297A (en) | System and method for realizing deep packet inspection technology based on multi-core and multi-thread | |
| CN103260007B (en) | Based on the intelligent monitor system of multiport storage controller on sheet | |
| CN102857703A (en) | High-definition video character superimposing system and control method | |
| CN106339484A (en) | System and method for intelligent video retrieval processing | |
| CN112449147A (en) | Video cluster monitoring system of photovoltaic power station and image processing method thereof | |
| CN116365717A (en) | High-voltage circuit safety monitoring system and method | |
| US20130322682A1 (en) | Profiling Activity Through Video Surveillance | |
| CN102148962A (en) | Dual-digital signal processing (DSP) visual tracking remote monitoring system | |
| CN103731316A (en) | Flow monitoring device and method | |
| CN109993286B (en) | Sparse neural network computing method and related product | |
| CN205385561U (en) | Tiled display systems of shielding more | |
| CN106557527A (en) | A kind of pornographic image detection and system | |
| Castillo et al. | Iot-based multi-view machine vision systems | |
| CN106908455A (en) | IMAQ terminal, glass surface defects on-line detecting system and method | |
| CN112291118A (en) | Multi-core data processing device and method based on FPGA | |
| CN110233683A (en) | AR edge calculations resource regulating method, system and medium | |
| CN204667397U (en) | A kind of movement circuit being applicable to the flow collection equipment of networks congestion control information audit | |
| CN203675196U (en) | Network control 3G-SDI high-definition characters superimposer | |
| CN109714569B (en) | Real-time fusion method and device for multi-channel video images | |
| CN102981809A (en) | Message-based non-synchronous 3D rendering method in multi-thread environment | |
| CN107276845A (en) | A kind of real time data dispatching method with fault tolerance of field of track traffic | |
| CN117896123A (en) | Security barrier system, electronic device, and computer storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |